Augusto Montisci

A Neural Network Diagnosis Approach for Analog Circuits

This paper presents a neural network system for the diagnosis of analog circuits and shows how the performance of such a system can be affected by the choice of different techniques used by its submodules. In particular we discuss the influence of feature extraction techniques such as Fourier Transforms, Wavelets and Principal Component Analysis. The system uses several different power supplies and as many neural networks “in parallel”. Two different algorithms that can be used to combine the candidate sets produced by each network are also presented. The system is capable of diagnosing multiple faults even if trained on single ones.

Neural network-based analog fault diagnosis using testability analysis

A fault diagnosis procedure for analog linear circuits is presented. It uses an off-line trained neural network as a classifier. The innovative aspect of the proposed approach is the way the information provided by testability and ambiguity group determination is exploited when choosing the neural network architecture. The effectiveness of the proposed approach is shown by comparing with similar work that has already appeared in the literature.

Geometrical synthesis of MLP neural networks

This paper presents a new constructive algorithm to design multilayer perceptron networks used as classifiers. The resulting networks are able to classify patterns defined in a real domain. The proposed procedure allows us to automatically determine both the number of neurons and the synaptic weights of networks with a single hidden layer. The approach is based on linear programming. It avoids the typical local minima problems of error back propagation and assures convergence of the method. Furthermore, it will be shown in this paper that the presented procedure leads to single-hidden layer neural networks able to solve any problem in classifying a finite number of patterns. The performances of the proposed algorithm have been tested on some benchmark problems, and they have been compared with those of different approaches.

A neural inverse problem approach for optimal design

An original approach to the optimization of electromagnetic structures is presented that makes use of a neural network trained to capture the functional relationship between the design parameters and the objective function. The algebraic structure of the network is used to find the basins of attraction of the objective function of the optimization problem, avoiding the major drawbacks of the commonly used algorithms, i.e., the entrapment in local minima, and/or the huge amount of cost function evaluations.

A signal-processing tool for non-destructive testing of inaccessible pipes

The design of Non-Destructive-Testing systems for fault detection in long and not accessible pipelines is an actual task in the industrial and civil environment. At this purpose, the diagnosis based on the propagation of guided ultrasonic waves along the pipes offers an attractive solution for the fault identification and classification. The authors studied this problem by means of suitable Artificial Neural Network models. Numerical techniques have been used to simulate the guided wave propagation in the pipes. In particular, the finite element method has been used to model different kinds of pipes and faults, and to obtain several returning echoes containing the faults information. Torsional wave modes have been used as excitation waves. The obtained signals have been processed in order to reduce the data dimensionality, and to extract suitable features. The features selected from the signals can be further processed in order to limit the size of the Neural Network models without loss of information. At this purpose, the principal component analysis has been investigated. Finally, the selected features have been used as input for the neural network models. In this paper, traditional feed-forward, multi-layer perceptron networks have been used to obtain the information on size and location of localized notches.

An automatic optical inspection system for the diagnosis of printed circuits based on neural networks

The aim of this work is to define a procedure to develop diagnostic systems for printed circuit boards, based on automated optical inspection with low cost and easy adaptability to different features. A complete system to detect mounting defects in the circuits is presented in this paper. A low-cost image acquisition system with high accuracy has been designed to fit this application. Afterward, the resulting images are processed using the wavelet transform and neural networks, for low computational cost and acceptable precision. The wavelet space represents a compact support for efficient feature extraction with the localization property. The proposed solution is demonstrated on several defects in different kind of circuits.

Inversion of MLP neural networks for direct solution of inverse problems

In this work, a neural-based approach for inverse problems in the field of electromagnetic devices design is presented. A multilayer perceptron neural network is first trained to solve the analysis problem of the studied system. As a design problem can be formulated as an inverse problem, i.e., starting from the design requirements the optimal values of the design parameters have to be obtained, the input of the neural network will correspond to the design parameters while the output is the objective function of the optimization problem. In this work, a procedure is presented which performs the inversion of the trained neural network when the design requirements are assigned to the output.

Multiobjective Tabu Search Algorithms for Optimal Design of Electromagnetic Devices

In this paper, an original algorithm to solve multiobjective optimization problems, which makes use of the tabu search meta-heuristic, is presented. Scalarization of the vector problem is performed by introducing fitness functions that take under control both the Pareto optimality of the solutions, and the uniformity in the Pareto front sampling. The performance of the proposed algorithm is compared with that of a scalar tabu search method, coupled with the -constraint strategy. The results on analytical and electromagnetic problems demonstrate the effectiveness of the method.

Acoustic Tomography for Non Destructive Testing of Stone Masonry

Acoustic methods seem to be very suitable for evaluating a building condition because they give information with immediacy, rapidity and relatively low cost. They are based on measurements of the velocity of acoustic waves propagating through the material. Those methods are often carried out applying the Through Transmission Technique, in which the wave is transmitted by a transducer through the test object and received by a second transducer on the opposite side. In this paper, the sonic tomography imaging based on this technique has been used for non destructive testing of a stone masonry. The tomography implies that a ill-posed linear equations system has to be solved, in order to determine the velocities distribution inside the tested structure, thus highlighting the presence of anomalies. Different inversion algorithms, chosen between the most commonly used, have been implemented for determining the distribution of waves velocity in selected sections of the tested masonry.

Artificial neural networks for non-destructive evaluation with ultrasonic waves in not accessible pipes

The design of non-destructive testing systems for fault detection in long and not accessible pipelines is an actual task in the industrial and civil environment. At this purpose the diagnosis based on the propagation of guided ultrasonic waves along the pipes offers an attractive solution for the fault identification and classification. The authors studied this problem by means of suitable artificial neural network models. Numerical techniques have been used to model different kinds of pipes and faults, and to obtain several returning echoes containing the fault information. These signals have been processed to filter the noise by using wavelets e blind separation methods and passed to a feature extractor system, whose purpose is to reduce the data dimensionality and to compute suitable features. The features selected from the signals have been further processed in order to limit the size of the neural network models without loss of information. At this purpose, the Garsons method and the principal component analysis have been investigated and compared. Finally, the extracted features have been used as input for the neural network models. In this paper, traditional feed-forward, multi layer perceptron networks have been used to classify position, width, and depth of the defects.